2024-12-10T13:48:47.576Z

<Tyler Stachecki> The # of upmaps is generally too large to assume a sorted vector I think (`std::find`. and `std::erase` are used in `apply_incremental` and would be O(n) with a sorted vector). It's possible that the size cost of the unordered map could be amortized somewhat for clusters that have the upmap balancer enabled, especially as PG sizes beyond 100/OSD become more commonplace.

2024-12-10T14:02:25.539Z

<Casey Bodley> somehow in <https://pad.ceph.com/p/rgw-weekly>, a bunch of text has been replaced with chinese(?). has anyone else had to deal with this? is there a way to revert that stuff?

2024-12-10T14:19:24.348Z

<gregsfortytwo> Looks like it made enough revisions the web UI doesn’t go far enough back in time, but somebody with database access can pull it out. David has done it before with CLT notes; perhaps one of the other admins can help

2024-12-10T15:19:43.668Z

<John Mulligan> This happens every so often on other pages too. I think a Chinese speaker visits the page and then has a tool do a translation and then that translated page somehow gets saved on the etherpad backend. It's rather annoying. I wish etherpad was a bit smarter and could prevent this kind of change (or at least force a confirmation dialog or something).

2024-12-10T16:55:28.923Z

<Casey Bodley> RFC for users of the ceph-backport script: <https://github.com/ceph/ceph/pull/60983>

2024-12-10T16:59:55.898Z

<Mr Kumar> Is any standard documents available for pvc storage cluster for aks platform?

2024-12-10T20:19:36.689Z

<Kyrylo Shatskyy> Casey, is it 100% reproducible in jenkins now?

2024-12-10T20:20:54.004Z

<Casey Bodley> no, just random failures. the last couple logs i checked ran on the same invingi05 node though

2024-12-10T20:21:16.246Z

<Casey Bodley> maybe something with the golang environment there?

2024-12-10T22:06:08.325Z

<Satoru Takeuchi> I guess this document is what you want.
<https://github.com/rook/rook/blob/master/deploy/examples/cluster-on-pvc.yaml>

2024-12-10T22:49:27.150Z

<Mark Nelson (nhm)> @stachecki.tyler I'd be curious how many we end up with on a typical cluster deployment and how we interact with it (repeated searches?) Memory fragmentation is a real issue for us.